The ideal statistical software for teaching would be powerful, easy to learn and free. Stata, SPSS and SAS are powerful and relatively easy to learn, but not exactly free. R is free and very powerful, but has a long-toed learning curve (see previous discussion on software package preferences here). At one point I toyed with PSPP (an open-source knock-off of SPSS) but it is a very poor substitute and not worth anyone's time. Now I've found Gretl (The GNU Regression, Econometric and Time-Series Library). Gretl is a stand-alone open-source cross-platform package. It can directly import files from Stata 9, Excel, csv and several other formats, although the imports do not always proceed smoothly. It uses menus or scripting, and it can invoke R for more sophisticated analysis and graphing (if R is installed). The built-in routines include the expected (e.g., ols and logit) as well as time-series and some arcana (3sls). The post-estimation analyses are first rate. Gretl won't supplant Stata on my personal desktop any time soon, but it's worth considering for the classroom if you are teaching statistics to law students with Excel or Datadesk.